Saving the Peccaries with the WWF
I've been asked for a software solution by a group from the World Wildlife Foundation, part of their Areas Project. Communication on this project has been slow because they're working out of the jungles of Peru, i.e. a days boat ride down the Amazon from any civilization or telephones. But the technical challenge is a good one, let me quote their explanation of the situation:
"We are working with white-lipped peccaries... and among other things we are trying to figure out how frequently these animals visit the salt licks, as well as how much time they spend there. The use of an mp3 recorder for this is based on the idea that by hooking it up to a receiver, we can record the pulses or beeps [emitted] by the radio collars that we put on the peccaries. Our setup consists of an antenna, mp3 recorder and a receiver set to scan thru the frequencies that belong to each of the collars we have put on the peccaries. So, if a peccary with a radio collar walks into the range of the antenna, the pulses of that particular collar will be picked up by the receiver and recorded on to the mp3 recorder. We also leave a radio collar in the range of the antenna, which acts as a marker, so when looking at the sound file, we know the start time of each cycle of the scan. So, what we need is a software [program] that can take the sound file, and turn it into data we can later analyze."
The intention is to analyze mp3's of 7-30 gigabytes. Their failed approach consisted of decoding these mp3's to wave, in order to feed it into a scientific program that serves this sort of purpose. I am ruling out converting these behemoths as any part of a reasonable solution, even if you got it to work it would probably cause hard drive failure.
So the approach is to take an open source mp3 decoder, but strip out the unnecessary process of writing to disk. Now we're decoding a mp3 into PCM data one frame at a time, which should be easy to analyze and detect something as simple as a beep. I'm building this upon the open source MPEG Audio Decoder or MAD. We're working so far with decent accuracy, and tested analyzing a 65 meg file in under 2 minutes.
"We are working with white-lipped peccaries... and among other things we are trying to figure out how frequently these animals visit the salt licks, as well as how much time they spend there. The use of an mp3 recorder for this is based on the idea that by hooking it up to a receiver, we can record the pulses or beeps [emitted] by the radio collars that we put on the peccaries. Our setup consists of an antenna, mp3 recorder and a receiver set to scan thru the frequencies that belong to each of the collars we have put on the peccaries. So, if a peccary with a radio collar walks into the range of the antenna, the pulses of that particular collar will be picked up by the receiver and recorded on to the mp3 recorder. We also leave a radio collar in the range of the antenna, which acts as a marker, so when looking at the sound file, we know the start time of each cycle of the scan. So, what we need is a software [program] that can take the sound file, and turn it into data we can later analyze."
The intention is to analyze mp3's of 7-30 gigabytes. Their failed approach consisted of decoding these mp3's to wave, in order to feed it into a scientific program that serves this sort of purpose. I am ruling out converting these behemoths as any part of a reasonable solution, even if you got it to work it would probably cause hard drive failure.
So the approach is to take an open source mp3 decoder, but strip out the unnecessary process of writing to disk. Now we're decoding a mp3 into PCM data one frame at a time, which should be easy to analyze and detect something as simple as a beep. I'm building this upon the open source MPEG Audio Decoder or MAD. We're working so far with decent accuracy, and tested analyzing a 65 meg file in under 2 minutes.
0 Comments:
Post a Comment
Links to this post:
Create a Link
<< Home